Headphone amp, Turn on pop [solved]

[Michaelaudio]

So ive build a mobile Headphone amp, which works good so far. On the Output i have the TS5A23166, that should be open durring the start up, to prevent Turn on pop. Even when the ctrl pins are pulled to Grund, i do have a verry loud Turn on pop. The Switch is normally open, so i shouldnt hear anything when powering Up, but i do and i have no Idea why.

Any suggestions?

[magic]

TS5A23166 appears to be a low voltage chip, are you sure that the initial transient doesn't exceed 5V, causing it to do something weird?
How long is the switch held open? You need to wait until all the transients in the whole signal path settle.

BTW, not sure what the purpose of U1A/U1B is supposed to be.

[Zero999]

It might be the output voltage of the amplifier is exceeding the power supply rails of the analogue switch.

Please post the full schematic and use the print screen button, rather than posting a big arse photograph of the monitor.

[Michaelaudio]

I'll post a pdf later when im Home.

The TS5A is supplied by a 7803. How ever i made a terrible mistake for the ctrl Input (way too high Voltage) BUT when i Just Connect it to ground its the same Problem. 

@magic U1 is an Input buffer, providing a low impedance source to the following stages.

Oh yes, damn. The OP amp supplie is 18V, biased with 9V on the Inputs, thats way to much for The TS5 ofc. This happens when you wanna do it "quick" bc it should be a Gift for someone.

Well i guess thats the Mistake, thanks Guys.

Would it be reliable to take a Switch that can handle the Voltage, and let it Switch on after ~2 seconds via a RC circuit?

[magic]

Those buffers drive high impedance noninverting inputs of subsequent opamps, so any difference they make is marginal.

Technically, the subsequent stage has closed loop gain, so its AC differential input voltage and capacitive input current is a little higher for the same input signal. But I don't know if the difference is enough to have a meaningful effect.

[Zero999]

The TS5A23166 is completely unsuitable. The signal from your circuit can drop below 0V, as well as go above 5V.

Use a small mechanical relay. A DPDT is ideal, as it can connect both the outputs to 0V, via a much lower resistors rapidly charging the capacitors, when they're disconnected from the headphones.

Here's an example of a suitable relay.
https://uk.rs-online.com/web/p/signal-relays/7181998

[Michaelaudio]

I more or less copied the Input Stage from Phils Lab HP amp Design, u may be right but i guess it wont make it worse

@Zero999 its a mobile amp, powered from a 9V 800maH battery, that would be too much current draw. Plus my Board dimensions cant be larger than 3x5 cm, so i have verry Limited space. It must be an ic not larger than SOIC8 footprint.

Or Just another solution, just that i have no Pop.

[magic]

You could free up some space by eliminating pointless voltage followers. Some signal relays are fairly small.

For the output, opamps with better load driving capability could be considered. Like NE5532. Or NJM4556A; beware of phase reversal with this one (very ugly negative clipping), but the preceding TL072 probably doesn't have enough output swing to trigger it. Your TL072 output drivers have phase reversal too.

And so does TL072 in the gain stage. You could reduce the amount and the ugliness of transients generated by that stage simply by using the top rail as ground (referring Vbias filter capacitor, C3 and output loads to it) and the bottom rail as a negative supply.

edit
I missed battery powered and low voltage. NE5532 at unity gain will only work to ~2V below the top rail, TL072 is similarly constrained at the bottom. Not a great choice; some 4558, LM883, MC34078 could give more, dunno about THD and power. NE5532 is a power hog and NJM4556A even more so, but you get performance you pay for. With the NJM one chip could drive both channels to over 50mA, but watch out for power dissipation. OTOH, maybe it's not too bad at 9V.

[Michaelaudio]

I'll Draw a new schematic and post it. But i cant See where i have a phase Shift, all signals are on the positive inputs of the op amps?

I'll look what i can find, maybe If i get rid of an OP it will fit.

[Zero999]

The pop can be minimised to a very low, almost inaudible level by using large capacitors for the bias voltage and input. Here's an example showing a 22mV output transient, with a slow rise time up to 500ms. It has the disadvantage of it taking few seconds to turn on and requiring physically larger parts, but they only have to be rated to 4.5V, since they only charge to half the supply voltage.

What's the impedance of the headphones? 47µF is probably a bit too low. I chose 470µF for plenty of bass. Again it only needs to be rated to 4.5V.

This schematic also shows the correct way to connect two op-amps in parallel. U1a is the master and U1b the slave.

[magic]

The other method works as well. It also provides 3dB reduction of output stage voltage noise, unlike yours. (Yours has 3dB current noise advantage, but source impedance is very low so it's irrelevant).

edit
Wait, no. I thought you are referring to running both amps at unity gain, driven by a separate gain stage.
But you combined gain stage and output, using the gain amp to supply half of load current. This works too, with somewhat more distortion, but less power consumption.

[Michaelaudio]

So thats the circuit now. I wanna use 2 5532 for lower distortion and Power handling. For the realay, the 072 should Work, since its coil draws only 11mA, for a few seconds.

47uF should be enough i guess, its about 13Hz with a 250ohm load, so way below any kind of frequencies in music.

For the PSU im not quite Sure. I have to Test it Out if 9V is enough, however i tested it with 15v, using a step up converter, thats nice for a stable supply Voltage, but the tiny step up i got from eBay, is noisy as hell. When i Connect the circuit to it, without Sound, its like FM when u have no Connection. Strangely, that completely is gone when music is playing (maybe bc of the load)

[Zero999]

Any switched mode converter will be noisy.

It will still make a pop, as the relay will be too slow to disconnect the headphones. It needs to be connected so they're normally disconnected and the relay activated, shortly after power up. That will mean extra power consumption, which I know is what you're trying to avoid. The op-amp's output voltage will be about 1.5V lower, than the supply voltage, so it might not switch the relay, at lower battery voltages.

What do the 10R resistors do? Are the headphones designed to be driven like that? Generally the output impedance of the amplifier should be as low as possible, to provide a good damping factor. Refer to the circuit I posted previously for the correct way to connect two op-amps in parallel and maintain a low output impedance.

Why not simply design it so there's no turn on pop? I've shown you one way with the previous schematic. Another way is to connect the headphones' 0V between two capacitors to either rail which should be closely matched.

Since you're trying to save space, there's little benefit in having a separate TL072 gain stage, followed by the NE5532 buffer. The gain is only 9.2, so the tiny reduction in distortion will be inaudible. You might use the output op-amp for the gain stage, as I showed previously.

[magic]

This looks good in LTspice but it will make nasty glitches if built with TL072 due to phase reversal. Use the top rail as ground.
edit: That version may also glitch if it's turned off and then immediately back on again, before bias filter cap discharges.

Neither TL072 nor NE5532 makes a good unity gain follower at low voltage, they clip asymmetrically some 2V away from one supply rail due to input stage limitations.

Regarding OP's circuit:
- C4 is too high, seems -3dB at 20kHz. Reduce it 10x.
- R6 could be removed, just increase R18/R19 to 10Ω each.
- Filter the bias to ground, making it simply 50% of the supply voltage is invitation for noise.
- Placing caps on opamp output typically achieves little benefit, certainly not filtering, may cause instability.

[Zero999]

This looks good in LTspice but it will make nasty glitches if built with TL072 due to phase reversal. Use the top rail as ground.
edit: That version may also glitch if it's turned off and then immediately back on again, before bias filter cap discharges.

You're right on both counts. The first won't be an issue with the NE5532, but the second will be a problem, which ever op-amp is used.

I was thinking, just couple the signal to both rails, with matched capacitors, that way there will be minimal pop, the voltage will all be pretty central, as it's powered up.

[magic]

This couples 50% of rail noise into the signal.
If the rail carries pulses corresponding to positive signal halves driven out by the output stage, those get coupled too.

Maybe some simple active circuitry could be devised to discharge the bias cap when the supply rail collapses.
BTW, it looks like the same glitch will also occur during shutdown, not only on subsequent startup.

edit
In some cases, phase reversal may be prevented by adding series resistance in front of the affected amplifier input. I recall seeing this working for a discrete PNP input stage somebody presented here. Add a diode clamping the capacitor to the supply rail and maybe it works out.

Another way is chips without phase reversal in the first place. NE5532 (but bipolar input) or OPA2134 maybe.
Although in some cases, opamps advertised as such may still have phase reversal if the inputs are taken far enough beyond the rails
This may also be a violation of absolute maximum ratings at the same time.

[Zero999]

This couples 50% of rail noise into the signal.
If the rail carries pulses corresponding to positive signal halves driven out by the output stage, those get coupled too.

It's being powered from a battery, which is very clean. If the pulses from the output stage are coupled to both rails, they should cancel.

[magic]

Not cancel, but add to create a complete, attenuated, inverted image of load current. Only in a dual supply situation, only if both rails have equal impedance (including reactance).

Cancellation would occur in case of cross-conduction in the output stage.

In singe supply configuration, you first draw current from the battery, pulling the rail down, and you charge the load coupling capacitor as you drive the load. Then you pull the capacitor to ground, using the stored charge to drive the load negative. Battery current is zero in the second phase, so rail voltage is unaffected. Hence the rail carries halfwave rectified signal.

BTW, TL072 has much worse PSRR from the negative rail, so arguably that's where ground should be. Fun

[Zero999]

Not cancel, but add to create a complete, attenuated, inverted image of load current. Only in a dual supply situation, only if both rails have equal impedance (including reactance).

Cancellation would occur in case of cross-conduction in the output stage.

In singe supply configuration, you first draw current from the battery, pulling the rail down, and you charge the load coupling capacitor as you drive the load. Then you pull the capacitor to ground, using the stored charge to drive the load negative. Battery current is zero in the second phase, so rail voltage is unaffected. Hence the rail carries halfwave rectified signal.

BTW, TL072 has much worse PSRR from the negative rail, so arguably that's where ground should be. Fun

What I mean by cancel is, each rail will have the half wave rectified signal superimposed on it. Par of the signal will be fed back to the input, but it will be inverted, as the output will pull the respective rail down slightly, so all it will do is reduce the total gain slightly, due to negative feedback.

The only downside I can see with this scheme is it will compromise the channel separation. One channel will couple to the other, via the power supply rails, but is it a big deal for a headphone amplifier? Everything is a compromise.

[jonovid]

I found this 

[wasedadoc]

That reverse biases the output electrolytic half the time.  Why is that capacitor there at all?

[Michaelaudio]

@Magic. Ur right about C4, R6. R6 should be a Shirt circuit Protection. I dont exactly know what u mean with "Filter bias to ground"
Caps on outputs: do u mean the Filter caps for the biasing, or the output for the relay Driver?
Should i just get rid of the parallel Ne55 and just switch the 72 to a 5532?

@Zero999 In Ur circuit the Input is shorted to ground, but thats Not the case in reality right?
U might be right about the relay, either i could Connect the 9V directly to the relay, and then switching a resistor for the 9V to the Rest of the circuit, or find a relay that only Takes 10mA. Well i have found some, but id like to find one that i can buy on a lokal Website, id rather not pay 20€ shipping costs for a 3€ relay.

The chanel Separation shouldnt be a big deal, i dont need -120db Crosstalk or something.

[Zero999]

@Zero999 In Ur circuit the Input is shorted to ground, but thats Not the case in reality right?

Shorting the input to ground was just a convenient way to model the turn on transient response.

I found this 

That's what I'm talking about. As mentioned above, the output capacitor is no longer required. It's better if the capacitors are matched.

[Michaelaudio]

Im not quite sure how that circuit works, and what is the purepose of the resistors across both Caps that go to ground/Vcc?

I found out that 9V isnt enough for my load, either i'll have to use a boost converter, or i simply Connect the OP amps in bridge Mode. I wouldnt need an output cap, would that already Eliminate the pop noise? II would think so.

[Zero999]

Im not quite sure how that circuit works, and what is the purepose of the resistors across both Caps that go to ground/Vcc?

I found out that 9V isnt enough for my load, either i'll have to use a boost converter, or i simply Connect the OP amps in bridge Mode. I wouldnt need an output cap, would that already Eliminate the pop noise? II would think so.

The resistors balance the DC voltages on the capacitors, as the leakage currents might differ. They also provide a path for the tiny current generated by the bias and offset from the op-amp.

I don't see how you can bridge, as headphones normally have a three way connectors.

Why not simply use two batteries?